FORUMS
Remove All Ads from XDA

Discussion thread for /data EMMC lockup/corruption bug

5,342 posts
Thanks Meter: 7,242
 
By sfhub, Senior Member on 9th May 2012, 01:08 PM
Post Reply Email Thread
7th June 2012, 07:05 AM |#481  
OP Senior Member
Thanks Meter: 7,242
 
More
Quote:
Originally Posted by Entropy512

There's no good way I am aware of to test a kernel for whether it is dangerous or not. Same for update-binaries.

You could whitelist the md5sum of known good binaries - but the whitelist will quickly grow monstrous and many "OK" binaries may get flagged.

Yes, that was the idea, update-binary white/black list rather than kernels.

The place I was thinking on adding it was try_update_binary() where the update-binary is already extracted for you. This would basically be modeled after signature verification (in that user could disable this feature if they REALLY wanted to install something). Also good/bad MD5s of update-binary would be kept in a file for easy edit without recompiles.

We only have a few update-binary executables on our phone that are used by most ROM developers. They don't change that often. All our GB updates had the same update-binary MD5 and I think all the ICS leaks had the same update-binary MD5. They are all listed in the first post under the post that says md5 in the description.

The ability to turn off the white/black list feature would be the catch-all to handle cases where the ROM developer used an unknown/unsafe update-binary.
The Following 2 Users Say Thank You to sfhub For This Useful Post: [ View ] Gift sfhub Ad-Free
 
 
7th June 2012, 02:48 PM |#482  
Member
Thanks Meter: 18
 
More
Quote:
Originally Posted by sfhub

The ability to turn off the white/black list feature would be the catch-all to handle cases where the ROM developer used an unknown/unsafe update-binary.

With all the late night work devs tend to do it is inevitable that someone will grab the wrong one.
8th June 2012, 12:58 AM |#483  
MR.change's Avatar
Senior Member
Flag Baghdad
Thanks Meter: 126
 
More
Quote:
Originally Posted by sfhub

Most folks are waiting for the source to drop. We've gotten hints our official ICS will be dropping pretty soon and obviously source will then be available.

sbrissen is working on kernel from intl gs2 update4. We are on CDMA so there are always changes on our platform from the intl version.

Even then there are many folks who actually prefer the stock kernel repacks.

Quote:
Originally Posted by Mattix724

Source code?

Sent from my SPH-D710 using xda premium

Quote:
Originally Posted by Entropy512

Because the Note is the first device (other than SHW-M250S/K/L) to receive official ICS (and source) that is affected by the bug.

All E4GT ICS releases so far are leaks - no source. And from the sounds of it, their official release (and source) should have the fix.

Oh well then, I do apologize for asking a stupid question in the first place , but I wasn't aware that the official ICS wasn't out yet for the E4GT
my bad, sorry
8th June 2012, 01:53 AM |#484  
OP Senior Member
Thanks Meter: 7,242
 
More
If you plan on running through this test plan, you should be able to deal with a superbrick (having TEP or within return period would be best). Through code analysis, it is believed these are all safe operations, but until we test, we don't know for sure. I would like to start with one volunteer, then expand from there if successful.

Downloads
a) SPH-D710.ODIN_FF02_KERNEL_CWM_AGAT.exe
b) update-BRICKFIXv1-cm-9-20120606-SNAPSHOT-epic4gtouch-alpha5-signed.zip
c) update-kernel-CM9a5.zip
d) SPH-D710.FF02_CL663858_ROOTED-oc-sfx.exe
e) AOKP_BRICKFIXv1_Build-37_epic4gtouch.zip
f) update-kernel-AOKP-37.zip
g) Format_Cache_Data-E4GT.zip

Pre-requisite
First, verify your EMMC is susceptible to the superbrick using this app from Chainfire
http://forum.xda-developers.com/show....php?t=1693704

Test Plan
0a) ODIN Flash FF02 Stock (SPH-D710.FF02_CL663858_ROOTED-oc-sfx.exe) (Power+VolDown)
0b) will automatically boot to stock FF02

1a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
1b) boot to recovery (Power+VolUp)
1c) Flash Format_Cache_Data-E4GT.zip
1d) Flash CM9a5 BRICKFIXv1
1e) reboot to CM9

2a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
2b) boot to recovery (Power+VolUp)
2c) Flash CM9a5 BRICKFIXv1
2d) reboot to CM9

3a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
3b) reboot to recovery (Power+VolUp)
3c) perform wipe data/factory reset
3d) Flash CM9a5 *kernel* update.zip
3e) reboot to CM9

4a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
4b) boot to recovery (Power+VolUp)
4c) Flash CM9a5 BRICKFIXv1
4d) reboot to CM9
4e) make change to home workspace like adding/removing an app

5a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
5b) reboot to recovery (Power+VolUp)
5c) Flash CM9a5 *kernel* update.zip
5d) perform nandroid backup
5e) perform wipe data/factory reset
5f) reboot to CM9
5g) confirm your change from 4e is not present

6a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
6b) reboot to recovery (Power+VolUp)
6c) perform nandroid restore
6d) reboot to CM9
6e) confirm your change from 4e is present

7) Repeat Steps 0-6 using AOKP-37 BRICKFIXv1 / kernel instead of CM9a5 BRICKFIXv1 / kernel

DISCLAIMER: you should understand that the core problem is a bug in the EMMC firmware. Unless that is "fixed", which is unlikely, you will ALWAYS be exposed to the superbrick to some extent. What we are doing is eliminating common paths you may reach the problem. If the workaround is placed in the kernel mmc driver, then all ROMs/Recoveries that use that kernel will have the call to the buggy EMMC firmware code disabled, however, you may unwittingly install a different kernel/recovery which doesn't have the mmc driver patch and expose yourself to the superbrick again. If you workaround the problem in the userspace utilities, you could always find a ROM which hasn't eliminated the format() call from the edify install script and is using an unsafe update-binary. Both workarounds are being actively pursued to reduce (but not eliminate) your chances of hitting the core EMMC firmware bug.

Results so far:
Esoteric68 - no issues - 6 flashes CM9 BRICKFIXv1, 3 flashes AOKP BRICKFIXv1, 3 wipe data/factory resets, 3 nandroid restores, 1 stock FF02 flash
http://forum.xda-developers.com/show...5#post27127645
http://forum.xda-developers.com/show...5#post27127645

krazy_smokezalot - no issues - Ran Test Plan twice - 6 flashes CM9 BRICKFIXv1, 6 flashes AOKP BRICKFIXv1, 4 wipe data/factory resets, 4 nandroid restores, 4 stock FF02 flash
http://forum.xda-developers.com/show...0#post27137440
http://forum.xda-developers.com/show...7#post27161837

dexter35803 - no issues - 4 flashes CM9 BRICKFIXv1, 3 flashes AOKP BRICKFIXv1, 6 wipe data/factory resets, 3 nandroid restores, 2 stock FF02 flash
http://forum.xda-developers.com/show...1#post27185711
The Following 4 Users Say Thank You to sfhub For This Useful Post: [ View ] Gift sfhub Ad-Free
8th June 2012, 02:23 AM |#485  
Esoteric68's Avatar
Senior Member
Flag Hellabama
Thanks Meter: 1,482
 
More
Quote:
Originally Posted by sfhub

If you plan on running through this test plan, you should be able to deal with a superbrick (having TEP or within return period would be best). Through code analysis, it is believed these are all safe operations, but until we test, we don't know for sure. I would like to start with one volunteer, then expand from there if successful.

Downloads
a) SPH-D710.ODIN_FF02_KERNEL_CWM_AGAT.exe
b) update-BRICKFIXv1-cm-9-20120606-SNAPSHOT-epic4gtouch-alpha5-signed.zip
c) update-kernel-CM9a5.zip

Preliminary Test Plan
1a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
1b) boot to recovery (Power+VolUp)
1c) Flash CM9a5 BRICKFIXv1
1d) reboot to CM9

2a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
2b) boot to recovery (Power+VolUp)
2c) Flash CM9a5 BRICKFIXv1
2d) reboot to CM9

3a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
3b) reboot to recovery (Power+VolUp)
3c) perform wipe data/factory reset
3d) Flash CM9a5 *kernel* update.zip
3e) boot to CM9

4a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
4b) boot to recovery (Power+VolUp)
4c) Flash CM9a5 BRICKFIXv1
4d) reboot to CM9
4e) make change to home workspace like adding/removing an app

5a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
5b) reboot to recovery (Power+VolUp)
5c) Flash CM9a5 *kernel* update.zip
5d) perform nandroid backup
5e) reboot to CM9

6a) ODIN AGAT FF02 CWM Repack (Power+VolDwn)
6b) reboot to recovery (Power+VolUp)
6c) perform nandroid restore
6d) reboot to CM9
6e) confirm your change from 4e is present

I'm game.
The Following 8 Users Say Thank You to Esoteric68 For This Useful Post: [ View ] Gift Esoteric68 Ad-Free
8th June 2012, 02:42 AM |#486  
OP Senior Member
Thanks Meter: 7,242
 
More
Quote:
Originally Posted by Esoteric68

I'm game.

That is great. I'll be around in case there are any questions.
The Following User Says Thank You to sfhub For This Useful Post: [ View ] Gift sfhub Ad-Free
8th June 2012, 02:50 AM |#487  
Esoteric68's Avatar
Senior Member
Flag Hellabama
Thanks Meter: 1,482
 
More
Quote:
Originally Posted by sfhub

That is great. I'll be around in case there are any questions.

After the reboot to CM9 on step 3 I may or may not be stuck. No brick just the CM9 boot animation is going for a longggggg time. Should I let it ride or pull the battery? I'm content to wait it out if needed.
The Following User Says Thank You to Esoteric68 For This Useful Post: [ View ] Gift Esoteric68 Ad-Free
8th June 2012, 02:59 AM |#488  
OP Senior Member
Thanks Meter: 7,242
 
More
Quote:
Originally Posted by Esoteric68

After the reboot to CM9 on step 3 I may or may not be stuck. No brick just the CM9 boot animation is going for a longggggg time. Should I let it ride or pull the battery? I'm content to wait it out if needed.

Did you flash the CM9a5 kernel, step 3d? CM9a5 won't boot with AGAT's kernel/CWM in place. It needs its own kernel.
The Following User Says Thank You to sfhub For This Useful Post: [ View ] Gift sfhub Ad-Free
8th June 2012, 03:20 AM |#489  
Esoteric68's Avatar
Senior Member
Flag Hellabama
Thanks Meter: 1,482
 
More
Quote:
Originally Posted by sfhub

Did you flash the CM9a5 kernel, step 3d? CM9a5 won't boot with AGAT's kernel/CWM in place. It needs its own kernel.

OMG Could XDA have chosen a worse time to go down? lol

Okay *deep breath* yes, I flashed the update kernel and this is the message I got.

assert failed: package_extract_file("zImage", "/tmp/zImage")
E:Error in /sdcard/update-kernel-CM9a5.zip
(Status 7)
Installation aborted

So thinking maybe I had a bad download I popped my sdcard out of the phone and re-downloaded the update kernel, went through the steps in #3 again and got the same message when I flashed the update kernel.

I wasn't supposed to extract the zImage myself, was I?

I'm currently chilling on CM9 while waiting for this forum to come back up.

So where do I go from here?
The Following User Says Thank You to Esoteric68 For This Useful Post: [ View ] Gift Esoteric68 Ad-Free
8th June 2012, 03:26 AM |#490  
OP Senior Member
Thanks Meter: 7,242
 
More
Quote:
Originally Posted by Esoteric68

OMG Could XDA have chosen a worse time to go down? lol

Okay *deep breath* yes, I flashed the update kernel and this is the message I got.

assert failed: package_extract_file("zImage", "/tmp/zImage")
E:Error in /sdcard/update-kernel-CM9a5.zip
(Status 7)
Installation aborted

So thinking maybe I had a bad download I popped my sdcard out of the phone and re-downloaded the update kernel, went through the steps in #3 again and got the same message when I flashed the update kernel.

I wasn't supposed to extract the zImage myself, was I?

I'm currently chilling on CM9 while waiting for this forum to come back up.

So where do I go from here?

EDIT: I figured out what was wrong with the kernel flash. I renamed the zImage by mistake. Will fix in a moment.

I would just reflash the full CM9 for now, while I figure out why that kernel flash failed.

Basically skip 3 and goto 4.
The Following User Says Thank You to sfhub For This Useful Post: [ View ] Gift sfhub Ad-Free
8th June 2012, 03:27 AM |#491  
Esoteric68's Avatar
Senior Member
Flag Hellabama
Thanks Meter: 1,482
 
More
Quote:
Originally Posted by sfhub

EDIT: I figured out what was wrong with the kernel flash. I renamed the zImage by mistake. Will fix in a moment.

I would just reflash the full CM9 for now, while I figure out why that kernel flash failed.

Basically skip 3 and goto 4.

Okay so then I'll wait lol
The Following User Says Thank You to Esoteric68 For This Useful Post: [ View ] Gift Esoteric68 Ad-Free
Post Reply Subscribe to Thread

Guest Quick Reply (no urls or BBcode)
Message:
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes